Why AMD videocards have so low overall score?

[gogamoga]

For example: Radeon HD5870 vs GTX 460

G3D Mark 2602 - 2674 (GTX 460 3% faster)

but 5870 clearly faster (http://www.hwcompare.com/8265/geforce-gtx-460-1gb-vs-radeon-hd-5870/ , http://www.anandtech.com/bench/product/511?vs=542) (5870 20% or more faster)

Radeon HD 7970 vs GTX 760

G3D Mark 5090 - 5007 ( 7970 1% faster)

but again (http://www.hwcompare.com/14814/geforce-gtx-760-vs-radeon-hd-7970/ , http://www.anandtech.com/bench/product/770?vs=854) (10% faster)

[David (PassMark)]

The hwcompare compare site is just poorly plagiarized data. The source data was from Tom's hardware, but they haven't accurately reported Tom's results.

For example, if you look at the actual source data for Bad Company,
http://www.tomshardware.com/charts/2010-gaming-graphics-charts-high-quality/Battlefield-Bad-Company-2,2466.html

The GTX 460 gets results between 33 and 68fps depending on the card model and setup.
The HD 5870 gets results between 51 and 92fps depending on the card model and setup.
So the actual results are not nearly as conclusive.

Similarly for Supreme Commander 2
The actual source data from Tom's shows no real difference in performance.
http://www.tomshardware.com/charts/2010-gaming-graphics-charts-high-quality/Supreme-Commander-2,2479.html

Different benchmarks and different games are going to favour different cards. In Arnand's charts there were several games where the GTX 460 was faster (Starcraft, Civ5, Batman, DirectCompute AES).

[gogamoga]

Different benchmarks and different games are going to favour different cards. In Arnand's charts there were several games where the GTX 460 was faster (Starcraft, Civ5, Batman, DirectCompute AES).

But this is rather an exception. Even 5850 faster than 460. The same applies to 7970 and 760.

[David (PassMark)]

You can't have one benchmark that exactly replicates the performance of a dozen different games, with multiple different resolutions, different quality settings, different operating systems and different device driver versions on different video cards. I could come up with a mix of games that replicates the PerformanceTest result for these two cards. ie. Show that they offer similar average performance across day 6 different games. But as you have pointed out it is equally possible to have a different mix of games, cards & settings that show one card is faster.

[gogamoga]

You can't have one benchmark that exactly replicates the performance of a dozen different games, with multiple different resolutions, different quality settings, different operating systems and different device driver versions on different video cards.

I just want to make a point that every single AMD video-card has their score lowered. There is no way when gtx 760 can be on par with 7970 except the app is optimized specially for Nvidia.

[David (PassMark)]

The story is the same for the gtx 760 to 7970 comparison.

The hwcompare.com link you posted in your first benchmark doesn't even have any relevant benchmarks. Unbelievably they have published a benchmark page without actually running a single benchmark. So it is completely meaningless.

The anandtech.com shows the cards are close (on average). If I pick Crysis, Civ 5 & Battlefield 3 as a suite of benchmarks then the gtx 760 is the winner. If I pick DiRT, Hitman & Sleeping Dogs then the 7970 is the winner.

Just because PerformanceTest behaves slightly more like Crysis and slightly less like DiRT doesn't mean it is wrong.

[gogamoga]

The story is the same for the gtx 760 to 7970 comparison.

The hwcompare.com link you posted in your first benchmark doesn't even have any relevant benchmarks. Unbelievably they have published a benchmark page without actually running a single benchmark. So it is completely meaningless.

T

I've posted hwcompare results just to point that 7970 has higher texture rate (118Gtexel/s vs 94.1), higher pixel rate (29.6 vs 23.5), wider memory bus (384 vs 256), while 760 has only slightly faster core and memory clock speed.


http://www.techpowerup.com/reviews/N...TX_760/27.html (comprehensive test)

[David (PassMark)]

The hwcompare site didn't measure the texture rate, etc... They really don't know what it is. If you read the page they actually say this is what the result might be in theory had they bothered to actually measure it. They really don't know (nor care) what the actual performance is.